Thursday, June 9, 2016

Cost of solar power (63)


The conservative side of politics in Australia has a goodly share of climate change deniers and fossil fuel proponents.  So it’s a delicious irony that the former leader of the Liberal Party (that’s the conservatives!), John Hewson, has now emerged as the Chairman of Solastor, a company that proposes to build a large solar thermal power station near Port Augusta.  RenewEconomy has details of the proposal (here, here and here).

I’ve written about the costs of a CST plant at Port Augusta before.  See here for details.

The Solastor proposal involves a modular heliostat/tower approach.  Each module will have a 24 m tower on which is sited a 10 tonne graphite block for thermal storage.  The footprint of the mirror field is 65 m × 35 m, for a total area of 2,300 m2.  We are told there will be approximately 100 heliostats per module, and I’m estimating the mirror area per module to be about 1,000 m2.

The thermal collection system then drives a conventional steam generator.  We are told only that the working steam temperature is 400°C and the storage temperature in graphite is 800°C.  We don’t know whether the condenser is water or air-cooled.

So this proposal is unconventional only in the sense of graphite as the storage medium.  Vast Solar are also aiming at a modular approach.

Before estimating the Levelised Cost of Electricity (LCOE) for the Solastor approach, let’s make a few back-of-envelope calculations about the performance using figures provided by RenewEconomy and Solastor.

Thermal collection:  Let me assume the optical efficiency of the field is 80%, the efficiency of the receiver is 95%, the mirror field area is 1,000 m2 and the DNI is 6.5 kWh per m2 per day.  Then the heat collected per module per day is 6.5 × 1,000 × 0.80 × 0.95 = 4,940 kWhth.  That agrees remarkably well with the figure given by Solastor in their slideshow presentation.

Thermal storage: The specific heat capacity of graphite is 0.71 kJ/kg.°C.  If the storage temperature range is 500°C, then the heat stored by a 10,000 kg graphite block is 10,000 × 500 × 0.71 kJth = 3,550 MJth = 0.986 MWhth.  That’s only about 1/3 of the thermal storage claimed by Solastor in their slideshow presentation.  I’m concerned that the thermal storage is under-specified for the requirements.

Solar multiple:  We are told in the RenewEconomy reports that Solastor proposes a 170 MW plant with 1,700 modules.  My estimate of the instantaneous collection capability of a 1,700 module system at 1 kW DNI per m2 is 1 × 1,700 × 1,000 × 0.80 × 0.95 = 1,292,000 kWth.  At a thermal to electric conversion efficiency of 33%, the possible power output would be 426 MWe, so the solar multiple of the proposed system is 426/170 = 2.5.

Cost and annual output:  According to figures provided by RenewEconomy, the 170 MW system to be completed in 2018 would cost AUD 1,200 million and have annual output 1,229,000 MWhe.  The Capacity Factor would be 1,229,000 / (365 × 24 × 170) = 0.83, which seems way too high to me (I would expect a figure of about 0.60), but let’s not quibble for the moment.

Let me now estimate the LCOE for the Solastor proposal using my standard assumptions:

  • there is no inflation,
  • taxation implications are neglected,
  • projects are funded entirely by debt,
  • all projects have the same interest rate (8%) and payback period (25 years), which means that the required rate of capital return is 9.4%,
  • all projects have the same annual maintenance and operating costs (2% of the total project cost), and
  • government subsidies are neglected.
For further commentary on my LCOE methodology, see posts on Real cost of coal-fired power, LEC – the accountant’s view, Cost of solar power (10) and (especially) Yet more on LEC.

Note that I am now using annual maintenance costs of 2% of capital cost rather than 3% as in posts during 2011. 

The results are as follows:
Cost per peak Watt              AUD 7.06/Wp
LCOE                                     AUD 112/MWh

The components of the LCOE are:
Capital           {0.094 × 1.2 × 109}/{1.229 × 106 MWh} = AUD 92/MWh
O&M              {0.020 × 1.2 × 109}/{1.229 × 106 MWh} = AUD 20/MWh

Conclusion

My LCOE estimates for comparable CST plants are Cerro Dominador (USD 121/MWh), the previous Alinta proposal for Port Augusta (AUD 218/MWh) and Atacama 1 (USD 149/MWh).   The Solastor LCOE would certainly be competitive with other CST plants provided Solastor could deliver on their performance and cost estimates.  However the team behind Solastor are not as experienced as other world-class CST players such as SolarReserve, so Solastor’s claims must be regarded with some suspicion until further details emerge.

This graphic has further details of the 63 LCOE estimates I have compiled.

Tuesday, May 10, 2016

Cost of solar power (62)


The PV world is buzzing with the recent announcement from the Emirate of Dubai about bids to construct the 800 MW Sheikh Maktoum Solar Park Phase III installation.  The winning bid was USD 30 per MWh; the under-bidder was USD 36.9 per MWh.  These bids are definitely under the cost of new-build fossil fuel power stations.

For information on the Sheikh Maktoum project, see here (RenewEconomy) and especially here (Apricum).

A couple of months ago, I reported on the Rubi PV project in Peru for which my estimate for the Levelised Cost of Electricity was USD 52/MWh.  Let’s see how the Sheikh Maktoum plant compares.

As is often the case, press reports for Sheikh Maktoum Solar Park Phase III do not give the hard information required for LCOE comparisons, namely peak power (AC to grid), capital cost and annual output.  All we have is 800 MW capacity and the winning bid: USD 30/MWh.

And as the linked Apricum report mentions …

“The price bid by Masdar/FRV [the winning bid for Sheikh Maktoum Solar Park III] is 19% lower than the second-lowest bid submitted by JinkoSolar.  It can be expected that both JinkoSolar and the third-lowest Acwa Power pushed their proposals very close to what can be considered commercially feasible today.  One may speculate how Masdar and FRV seemingly manage to play in a universe of their own.  Because the majority of the expenses for a solar plant lie in the upfront cost of construction, which gets recovered over numerous years, the cost of financing is a key overall cost driver.  One can suspect that Masdar had access to long-term financing through the wealthy emirate of Abu Dhabi that no commercial banks, the primary source of capital for the other bidders, could match in cost.

Bingo!  I reckon that hits the nail on the head.

The Apricum report also goes on to speculate (knowledgeably it seems to me!) that the winning bid incorporated one-axis tracking and that the capital cost for a 800 MW AC facility would be around USD 1.0 billion, or USD 1.25/Wp.

[For comparison purposes, my previous blog post contained reliable costs for 22 recent bids for government supported PV projects in Australia.  The average capital cost (AC to grid) was AUD 2.25/Wp, equivalent to USD 1.67/Wp at today’s conversion rate.  I expect things could be done a bit cheaper in Dubai, so let’s stick with Apricum’s cost estimate.]

What about the annual output of PV plants with one-axis tracking in Dubai?  My previous post referred to data from ARENA (Australian Renewable Energy Agency).  Out of the 22 projects that were described, the average Capacity Factor (AC to grid) for one-axis systems was 26%, and the CF figure for systems in the state of Queensland, probably a similar solar resource to Dubai, was 28%.  Let’s use that figure.

So for Sheikh Maktoum Solar Park III, we could estimate the annual output to be 24×365×0.28×800 = 1,962,240 MWh.

Let me now estimate the LCOE for Sheikh Maktoum Solar Park III using my standard assumptions:
  • there is no inflation,
  • taxation implications are neglected,
  • projects are funded entirely by debt,
  • all projects have the same interest rate (8%) and payback period (25 years), which means that the required rate of capital return is 9.4%,
  • all projects have the same annual maintenance and operating costs (2% of the total project cost), and
  • government subsidies are neglected.
For further commentary on my LCOE methodology, see posts on Real cost of coal-fired power, LEC – the accountant’s view, Cost of solar power (10) and (especially) Yet more on LEC.

Note that I am now using annual maintenance costs of 2% of capital cost rather than 3% as in posts during 2011.

The results are as follows:
Cost per peak Watt              USD 1.25/Wp
LCOE                                     USD 58/MWh

The components of the LCOE are:
Capital           {0.094 × 109}/{1.962 × 106 MWh} = USD 48/MWh
O&M              {0.020 × 109}/{1.962 × 106 MWh} = USD 10/MWh

Conclusion

The estimate using my standard methodology is nearly double the price bid by Masdar/FRV, USD 58/MWh compared to USD 30/MWh.  I think my maintenance costs are a bit high, but the biggest explanatory factor would be the financing costs.  I’m sure that the proponents for Sheikh Maktoum Solar Park III have financing costs that are much less than 8% per annum over 25 years.  There may also be taxation advantages that are specifically excluded in my methodology.

For comparison of these costs with installations around the world, please see my LCOE graphic (which I need to update).

One last comment.  The general trend is clear: the price of solar is coming down, the price of fossil fuel and nuclear generation is going up.  Moreover, the crossover point has probably already been reached, and if a realistic cost of carbon emissions were included then it would be game over for fossil fuel generators.

Tuesday, March 8, 2016

Cost of solar power (61)


ARENA (Australian Renewable Energy Agency) has recently released data on 22 projects that are applying for government funds to build large PV installations in Australia.  RenewEconomy has more details.  This is a treasure trove of information, and ARENA is to be congratulated for the initiative and its diligence.  The 22 proponents have now been invited to prepare complete applications for financial support from ARENA. 

By the way, the Australian Government is still trying to close down ARENA as a separate agency (in favour of moving it under the direct control of the Environment Minister), and they also are still trying to repeal laws governing the Clean Energy Finance Corporation.  So far, the government has been thwarted by cross-benchers in the Senate, but an election is due this year.  And the world is still getting hotter and hotter.

But, back to estimates for the Levelised Cost of Electricity (LCOE) …

The 22 projects are all multi-megawatt, come from all states in Australia and include both fixed and tracking systems.  There are some interesting variations between states and fixed/tracking technologies, but I won’t dwell on those.  Rather I’ll just look at average figures for the 22 projects.  This will give a comprehensive snapshot of costs in Australia as of today.

As stated, the average Capacity Factors across all 22 projects were 0.202 (DC ex-panels to inverter) and 0.246 (AC ex-inverter to grid).  That’s a bit confusing, but means the average proponent undersized the inverters relative to the panels by about 10%, presumably as a result of their costing calculations for optimal plant performance.  I’ll calculate the annual output of the average 1 kW system in AC to the grid as 0.202 × 24 × 365 = 1,770 kWh.

What about capital cost?  Well the ARENA data is helpful there too.  The average costs are AUD 1.86 per Watt (DC ex-panels) and AUD 2.25 per Watt (AC to grid).

Let me now estimate the LCOE using my standard assumptions:
  • there is no inflation,
  • taxation implications are neglected,
  • projects are funded entirely by debt,
  • all projects have the same interest rate (8%) and payback period (25 years), which means that the required rate of capital return is 9.4%,
  • all projects have the same annual maintenance and operating costs (2% of the total project cost), and
  • government subsidies are neglected.
For further commentary on my LCOE methodology, see posts on Real cost of coal-fired power, LEC – the accountant’s view, Cost of solar power (10) and (especially) Yet more on LEC. 

Note that I am now using annual maintenance costs of 2% of capital cost rather than 3% as in posts during 2011. 

The results for the average of the 22 ARENA PV proposals are as follows:

Cost per peak Watt              AUD 2.25/Wp
LCOE                                     AUD 144/MWh

The components of the LCOE are:
Capital           {0.094 × 2,250}/{1.770 MWhr} = AUD 119/MWhr
O&M              {0.020 × 2,250}/{1.770 MWhr} = AUD 25/MWhr

Conclusion

According to my standard methodology, the LCOE for the average of the proposed 22 large PV projects is AUD 144 per MWh.  This is rather higher than most LCOE students would calculate.  As Giles Parkinson of RenewEconomy writes:

“As a rough guide, a capital cost of around $2.20/watt AC is expected to translate into a levelised cost of generation of around $125/MWh, based on ARENA’s 10 per cent weighted average cost of capital methodology.”

However, in my defence, I’d say that my methodology does not include taxation advantages due to depreciation, which are included in the ARENA methodology.  I specifically exclude such considerations because I’m interested in international comparisons, and tax laws vary greatly between countries.

For comparison of these current Australian costs with installations around the world, please see my LCOE graphic (which I need to update).

One last point.  The ARENA data also includes details of operating expenditure.  For some time now, I have been thinking that my annual O&M costs (2% of capital cost) are high.  The ARENA data will allow me to check that.  But that will be the topic of my next post.

Monday, February 29, 2016

Cost of solar power (60)


There was an interesting article by Henry Lindon in Sustainnovate recently about record low PV prices that were bid into a renewable energy solicitation in Peru.  The winning bid by Enel Green Power was USD 47.98 per MWh for a 144 MW project; the second successful bidder was Enersur at USD 48.50 per MWh for a 40 MW installation.

In the same context, this link to RenewEconomy shows other recent competitive PV bids in PV reverse auctions.  There are many successful bids in the range USD 50 to USD 80 per MWh.

This leaves me in a quandary.  How to relate these low bids to my analysis of the Levelised Cost of Electricity?  Who knows what financial engineering underlies the bids?  Who knows whether it’s a loss-leader, or the recipient of a shadowy export enhancement grant?

The only hope is to find more complete specifications for these projects in the hope that there will be sufficient information for me to apply my standard LCOE analysis.  Fortunately there might be just enough information available from these recent events in Peru.  Let me explain …

The Rubi PV project by Enel Green Power is to be located in the Mocquegua region of southern Peru, which Wikipedia informs me is at latitude 17°S and elevation 1,410 m.  I’m sure the solar resource would be outstanding.  This press release describes the successful PV bid and includes details of Enel’s associated successful wind and hydropower bids in Peru.

The Rubi PV project has peak power 144 MW AC and annual output 440 HWh, whilst the 126 MW Nazca wind project will generate 600 GWh per year and the 20 MW Ayanunga hydro project will generate 140 GWh per year.  Taken together as a single entity, the peak power output is 290 MW and the annual output is 1,180 GWh.  The respective Capacity Factors are 0.34 (solar PV), 0.54 (wind) and 0.80 (hydro).  The respective CO2 abatements are 270 kt (PV), 370 kt (wind) and 109 kt (hydro).

The cost of the three projects is approximately USD 400 million and the installations will be complete before 2018.  That’s it – the full extent of the information.

Let me make a pro rata calculation for the cost of the PV component, that is assume that the cost of the Buri PV project is 144 MW × USD 400 million / 290 MW = USD 198.6 million.

I’ll now analyse the Buri PV project using my standard assumptions:
  • there is no inflation,
  • taxation implications are neglected,
  • projects are funded entirely by debt,
  • all projects have the same interest rate (8%) and payback period (25 years), which means that the required rate of capital return is 9.4%,
  • all projects have the same annual maintenance and operating costs (2% of the total project cost), and
  • government subsidies are neglected.
For further commentary on my LCOE methodology, see posts on Real cost of coal-fired power, LEC – the accountant’s view, Cost of solar power (10) and (especially) Yet more on LEC. 

Note that I am now using annual maintenance costs of 2% of capital cost rather than 3% as in posts during 2011. 

The results for the Buri PV installation are as follows:

Cost per peak Watt              USD 1.38/Wp
LCOE                                     USD 52/MWh

The components of the LCOE are:

Capital           {0.094 × 199×106}/{440,000 MWhr} = USD 43/MWhr
O&M              {0.020 × 199×106}/{440,000 MWhr} = USD 9/MWhr

Conclusion

Well, those numbers are astonishing, both for cost per peak Watt and LCOE.  Admittedly this part of Peru seems to be a paradise for solar energy, but this LCOE estimate is the best I have ever seen, and by a long way.  It’s very similar to the price that Enel Green Power successfully bid into the solicitation.

I fully realise that my pro rata calculation of the PV component is based on a heroic assumption, but at least that has to be my first approximation.  (It does mean that the wind and hydro components of Enel's offering will be very cheap indeed.)

The closest comparison that I have analysed is the 100 MW Amanecer PV project in Chile, also located in a region with a superb solar resource.  For Amanecer, I estimated USD 2.61 per peak Watt and USD 110 per MWh.

I’m shaking my head in wonder.  I just don’t see how Enel Green Power can do things at half the cost of Amanecer.

Acknowledgement: Thanks to Anthony Kitchener for referring me to the Rubi PV project.

Thursday, January 28, 2016

Cost of BIPV (1)


There was an interesting story in One Step Off The Grid this week about a PV installation at 101 Collins St, in the heart of Melbourne’s CBD.  This building is a 56 storey skyscraper and the 180 330 W PV panels are mounted vertically on the roof of the building at a height of 195 m.  The roof space is minimal, hence the vertical installation, and hence I’m going to classify this as Building Integrated Photovoltaics, BIPV.

The power from the installation will be used to offset the electricity demand of the building’s cooling system, so this is a behind-the-meter application.

The specifications of the installation are clearly stated in the article mentioned above.  The peak output of the system is 59.4 kW (I’m presuming at grid AC), the annual output will be 47 MWh, which is said to offset 59 tonnes of CO2 emissions per year.  (I’ll mention that most of Melbourne’s electricity supply is powered by brown coal, which has horrendous CO2 emissions.  In this case the data suggests, 59/47 t =1.26 t CO2 per MWh.)  The cost of the installation is stated as AUD 230,000.

The Capacity Factor of the installation is 47,000/(59.4×365×24) = 0.09, which is the lowest value I have recorded in all my studies.  The poor CF results from a combination of the vertical installation and Melbourne’s climate (not that I wish to offend my Melbourne friends).

Let me now estimate the LCOE using my standard assumptions:
  • there is no inflation,taxation implications are neglected,
  • projects are funded entirely by debt,
  • all projects have the same interest rate (8%) and payback period (25 years), which means that the required rate of capital return is 9.4%,
  • all projects have the same annual maintenance and operating costs (2% of the total project cost), and
  • government subsidies are neglected.
For further commentary on my LCOE methodology, see posts on Real cost of coal-fired power, LEC – the accountant’s view, Cost of solar power (10) and (especially) Yet more on LEC.

Note that I am now using annual maintenance costs of 2% of capital cost rather than 3% as in posts during 2011.

The results for the 101 Collins St installation are as follows:
Cost per peak Watt              AUD 3.87/Wp
LCOE                                     AUD 558/MWh

The components of the LCOE are:
Capital           {0.094 × 230,000}/{47 MWh} = AUD 460/MWh
O&M              {0.020 × 230,000}/{47 MWh} = AUD 98/MWh

Conclusion

At AUD 558 per MWh, the LCOE for this project is about 4 times that of utility-scale PV projects I have analysed recently.  For comparisons, I refer you to my LCOE graphic.

As mentioned, I’m going to classify this as BIPV.  The results are completely unrepresentative of the LCOE for best practice PV installations around the world.

Wednesday, December 30, 2015

PV versus CST


2015 was a watershed year.

The Paris climate change talks revealed a positive change in attitude amongst many countries (even if the signed agreement is rather toothless), steaming coal consumption has entered a structural decline, China is getting serious about air pollution, renewable electricity generation is cheaper than from fossil fuels if conditions are favourable, and battery storage is much talked about for numerous applications.

And it’s batteries that I want to blog about today.

Many studies on this blog and elsewhere show that PV is cheaper than Concentrated Solar Thermal (CST) power generation in the absence of storage.  But has battery technology advanced so much that PV plus batteries can compete with CST plus thermal storage at utility scale?  That’s the question I’ll answer.

In 2015 I analysed two installations in the Atacama desert where the solar conditions are superb.  The first was a PV installation at Amanecer, the second was the Atacama 1 CST project.  These are big projects delivered at world’s best practice. 

The Amanecer project had peak power 100 MW, no storage, Capacity Factor 0.308, annual output of 270 GWh, total cost USD 260.5 million, and my LCOE estimate was USD 110/MWh.  The Atacama 1 CST project was a conventional heliostat-tower design with twin tank molten salt energy storage for 17.5 hours, peak power 110 MW, estimated annual output 840 GWh, total cost USD 1.1 billion, and my LCOE estimate was USD 149/MWh.

What would happen if we tried to replicate the output of Atacama 1 with PV plus batteries?  Let me use the following assumptions:

  • PV costs and output are as per my Amanecer blog post,
  • batteries have a round-trip efficiency of 95% for a charge/discharge cycle,
  • batteries last 12.5 years under a regime with a complete charge/discharge cycle each day to a depth of 70%, and
  • the capital cost of batteries lies in the range USD 100 to USD 400 per kWh.
Suppose we want to replicate the peak power of Atacama 1 with PV, namely 110 MW.  Such a PV system would produce (110/100) × 270 = 297 GWh per year.  To match the annual output of Atacama 1, namely 840 GWh, requires that 840 - 297 = 543 GWh be delivered via batteries, or that 543 / 0.95 = 571.6 GWh be delivered by PV panels after accounting for the round-trip efficiency of storage.  Since the Capacity Factor for the site is 0.308, the peak power of the panels would be 571.6 / (0.308 × 24 × 365) = 0.212 GW or 212 MW.  The cost of those panels would be (212/100) × 260.5 = USD 552 million.

What about the cost of the batteries?

Well, we need to deliver 543 GWh annually, or 1,487,671 kWh per day.  But the batteries are assumed good for 70% discharge on a daily basis, so we need storage of 1,487,671/0.7 = 2,125,244 kWh.

And there’s more … We know the PV panels will last for 25 years, whereas the batteries are assumed to last for only 12.5 years.  So we need two sets of batteries during the assumed 25 year life of the project.  That makes 2 × 2,125,244 = 4,250,489 kWh battery storage required.

Exploring the sensitivity, the total cost of the batteries will be:
  • USD 425 million at battery cost USD 100 per kWh
  • USD 850 million at battery cost USD 200 per kWh
  • USD 1,275 million at battery cost USD 300 per kWh
  • USD 1,700 million at battery cost USD 400 per kWh
All up, to replicate the Atacama 1 CST project with PV plus batteries we need to add USD 260.5 million plus USD 552 million plus the cost of the batteries given above.  That makes:
  • USD 1.238 billion at battery cost USD 100 per kWh
  • USD 1.622 billion at battery cost USD 200 per kWh
  • USD 2.088 billion at battery cost USD 300 per kWh
  • USD 2.512 billion at battery cost USD 400 per kWh
Those figures need to be compared with the USD 1.1 billion cost of the Atacama 1 CST plant. 

Note that the capital price in 2015 for batteries is around USD 350 per kWh, so I think the result is clear.  Batteries are already cost efficient for portable electronic devices, maybe break even with flywheel costs for short-term frequency control, and in a few years will be cost-efficient for behind-the-meter applications and automobiles.  However the above estimates show that battery storage is nowhere near competitive with CST plus thermal storage for utility-scale applications.  In my view, proponents of utility-scale CST with storage can proceed with confidence.

With that (and as an enthusiastic inventor and developer of CST concepts), I wish all readers of this blog a successful and happy year ahead.  Thank you for reading this blog.

Acknowledgement: Thanks to Anthony Kitchener for suggesting that I perform this analysis.